Log10-Transform the Assay Response (mask-aware)

Applies log10() to the response column when is_log_response is TRUE. Non-positive values are floored to an adaptive minimum before transform.

Usage

compute_log_response(
  data,
  response_variable,
  is_log_response = TRUE,
  include_col = "included",
  floor_value = NULL,
  floor_method = "adaptive",
  verbose = FALSE
)

Arguments

data

Data frame.

response_variable

Character. Name of the response column.

is_log_response

Logical. Apply log10? Default TRUE.

include_col

Character. Name of the logical column marking rows that entered the fit (TRUE) versus masked rows (FALSE). If the column is absent, every row is treated as included (backward compatible).

floor_value

Numeric or NULL. If supplied, this exact floor is used for non-positive values instead of deriving one. Used to share a single floor across the standards and blanks frames.

floor_method

Character. How to derive the floor when floor_value is NULL: "adaptive" (default) uses 1\ positive value; "fixed" uses 1e-6.

verbose

Logical. Emit messages about floored values.

Value

data with the response column optionally transformed. The floor actually used is attached as attr(., "response_floor").

Details

The adaptive floor is a set-level statistic: it is derived from the included positive responses only (data[included & value > 0]), but the floor and the log10() are then applied to all rows. This is what lets masked points land on the same response axis as the fitted points without influencing the transform (see preprocess_standards()). Pass an explicit floor_value to reuse a floor computed elsewhere (e.g. to transform blanks on the same axis as the standards).